Eliminate use of curthread in if_ioctl functions by passing down the
[dragonfly.git] / sys / netinet6 / icmp6.c
1 /*      $FreeBSD: src/sys/netinet6/icmp6.c,v 1.6.2.13 2003/05/06 06:46:58 suz Exp $     */
2 /*      $DragonFly: src/sys/netinet6/icmp6.c,v 1.8 2004/03/06 05:00:41 hsu Exp $        */
3 /*      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $ */
4
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *      The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *      This product includes software developed by the University of
49  *      California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
67  */
68
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 #include <sys/domain.h>
84
85 #include <net/if.h>
86 #include <net/route.h>
87 #include <net/if_dl.h>
88 #include <net/if_types.h>
89
90 #include <netinet/in.h>
91 #include <netinet/in_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet/icmp6.h>
95 #include <netinet6/mld6_var.h>
96 #include <netinet/in_pcb.h>
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/nd6.h>
99 #include <netinet6/in6_ifattach.h>
100 #include <netinet6/ip6protosw.h>
101
102 #ifdef IPSEC
103 #include <netinet6/ipsec.h>
104 #include <netproto/key/key.h>
105 #endif
106
107 #ifdef FAST_IPSEC
108 #include "ipsec.h"
109 #include <netipsec/key.h>
110 #define IPSEC
111 #endif
112
113 #include <net/net_osdep.h>
114
115 #ifdef HAVE_NRL_INPCB
116 /* inpcb members */
117 #define in6pcb          inpcb
118 #define in6p_laddr      inp_laddr6
119 #define in6p_faddr      inp_faddr6
120 #define in6p_icmp6filt  inp_icmp6filt
121 #define in6p_route      inp_route
122 #define in6p_socket     inp_socket
123 #define in6p_flags      inp_flags
124 #define in6p_moptions   inp_moptions6
125 #define in6p_outputopts inp_outputopts6
126 #define in6p_ip6        inp_ipv6
127 #define in6p_flowinfo   inp_flowinfo
128 #define in6p_sp         inp_sp
129 #define in6p_next       inp_next
130 #define in6p_prev       inp_prev
131 /* macro names */
132 #define sotoin6pcb      sotoinpcb
133 /* function names */
134 #define in6_pcbdetach   in_pcbdetach
135 #define in6_rtchange    in_rtchange
136
137 /*
138  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
139  * others...
140  */
141 #define in6p_ip6_nxt    inp_ipv6.ip6_nxt
142 #endif
143
144 /*
145  * ppratecheck() is available, so define it here.
146  */
147 #define HAVE_PPSRATECHECK
148
149 extern struct domain inet6domain;
150 extern struct ip6protosw inet6sw[];
151 extern u_char ip6_protox[];
152
153 struct icmp6stat icmp6stat;
154
155 extern struct inpcbinfo ripcbinfo;
156 extern int icmp6errppslim;
157 static int icmp6errpps_count = 0;
158 static struct timeval icmp6errppslim_last;
159 extern int icmp6_nodeinfo;
160
161 static void icmp6_errcount (struct icmp6errstat *, int, int);
162 static int icmp6_rip6_input (struct mbuf **, int);
163 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
164 static const char *icmp6_redirect_diag (struct in6_addr *,
165         struct in6_addr *, struct in6_addr *);
166 #ifndef HAVE_PPSRATECHECK
167 static int ppsratecheck (struct timeval *, int *, int);
168 #endif
169 static struct mbuf *ni6_input (struct mbuf *, int);
170 static struct mbuf *ni6_nametodns (const char *, int, int);
171 static int ni6_dnsmatch (const char *, int, const char *, int);
172 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
173                           struct ifnet **, char *);
174 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
175                                 struct ifnet *, int);
176 static int icmp6_notify_error (struct mbuf *, int, int, int);
177
178 #ifdef COMPAT_RFC1885
179 static struct route_in6 icmp6_reflect_rt;
180 #endif
181
182
183 void
184 icmp6_init()
185 {
186         mld6_init();
187 }
188
189 static void
190 icmp6_errcount(stat, type, code)
191         struct icmp6errstat *stat;
192         int type, code;
193 {
194         switch (type) {
195         case ICMP6_DST_UNREACH:
196                 switch (code) {
197                 case ICMP6_DST_UNREACH_NOROUTE:
198                         stat->icp6errs_dst_unreach_noroute++;
199                         return;
200                 case ICMP6_DST_UNREACH_ADMIN:
201                         stat->icp6errs_dst_unreach_admin++;
202                         return;
203                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
204                         stat->icp6errs_dst_unreach_beyondscope++;
205                         return;
206                 case ICMP6_DST_UNREACH_ADDR:
207                         stat->icp6errs_dst_unreach_addr++;
208                         return;
209                 case ICMP6_DST_UNREACH_NOPORT:
210                         stat->icp6errs_dst_unreach_noport++;
211                         return;
212                 }
213                 break;
214         case ICMP6_PACKET_TOO_BIG:
215                 stat->icp6errs_packet_too_big++;
216                 return;
217         case ICMP6_TIME_EXCEEDED:
218                 switch (code) {
219                 case ICMP6_TIME_EXCEED_TRANSIT:
220                         stat->icp6errs_time_exceed_transit++;
221                         return;
222                 case ICMP6_TIME_EXCEED_REASSEMBLY:
223                         stat->icp6errs_time_exceed_reassembly++;
224                         return;
225                 }
226                 break;
227         case ICMP6_PARAM_PROB:
228                 switch (code) {
229                 case ICMP6_PARAMPROB_HEADER:
230                         stat->icp6errs_paramprob_header++;
231                         return;
232                 case ICMP6_PARAMPROB_NEXTHEADER:
233                         stat->icp6errs_paramprob_nextheader++;
234                         return;
235                 case ICMP6_PARAMPROB_OPTION:
236                         stat->icp6errs_paramprob_option++;
237                         return;
238                 }
239                 break;
240         case ND_REDIRECT:
241                 stat->icp6errs_redirect++;
242                 return;
243         }
244         stat->icp6errs_unknown++;
245 }
246
247 /*
248  * Generate an error packet of type error in response to bad IP6 packet.
249  */
250 void
251 icmp6_error(m, type, code, param)
252         struct mbuf *m;
253         int type, code, param;
254 {
255         struct ip6_hdr *oip6, *nip6;
256         struct icmp6_hdr *icmp6;
257         u_int preplen;
258         int off;
259         int nxt;
260
261         icmp6stat.icp6s_error++;
262
263         /* count per-type-code statistics */
264         icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
265
266 #ifdef M_DECRYPTED      /*not openbsd*/
267         if (m->m_flags & M_DECRYPTED) {
268                 icmp6stat.icp6s_canterror++;
269                 goto freeit;
270         }
271 #endif
272
273 #ifndef PULLDOWN_TEST
274         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
275 #else
276         if (m->m_len < sizeof(struct ip6_hdr)) {
277                 m = m_pullup(m, sizeof(struct ip6_hdr));
278                 if (m == NULL)
279                         return;
280         }
281 #endif
282         oip6 = mtod(m, struct ip6_hdr *);
283
284         /*
285          * Multicast destination check. For unrecognized option errors,
286          * this check has already done in ip6_unknown_opt(), so we can
287          * check only for other errors.
288          */
289         if ((m->m_flags & (M_BCAST|M_MCAST) ||
290              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
291             (type != ICMP6_PACKET_TOO_BIG &&
292              (type != ICMP6_PARAM_PROB ||
293               code != ICMP6_PARAMPROB_OPTION)))
294                 goto freeit;
295
296         /* Source address check. XXX: the case of anycast source? */
297         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
298             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
299                 goto freeit;
300
301         /*
302          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
303          * don't do it.
304          */
305         nxt = -1;
306         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
307         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
308                 struct icmp6_hdr *icp;
309
310 #ifndef PULLDOWN_TEST
311                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
312                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
313 #else
314                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
315                         sizeof(*icp));
316                 if (icp == NULL) {
317                         icmp6stat.icp6s_tooshort++;
318                         return;
319                 }
320 #endif
321                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
322                     icp->icmp6_type == ND_REDIRECT) {
323                         /*
324                          * ICMPv6 error
325                          * Special case: for redirect (which is
326                          * informational) we must not send icmp6 error.
327                          */
328                         icmp6stat.icp6s_canterror++;
329                         goto freeit;
330                 } else {
331                         /* ICMPv6 informational - send the error */
332                 }
333         } else {
334                 /* non-ICMPv6 - send the error */
335         }
336
337         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
338
339         /* Finally, do rate limitation check. */
340         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
341                 icmp6stat.icp6s_toofreq++;
342                 goto freeit;
343         }
344
345         /*
346          * OK, ICMP6 can be generated.
347          */
348
349         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
350                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
351
352         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
353         M_PREPEND(m, preplen, M_DONTWAIT);
354         if (m && m->m_len < preplen)
355                 m = m_pullup(m, preplen);
356         if (m == NULL) {
357                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
358                 return;
359         }
360
361         nip6 = mtod(m, struct ip6_hdr *);
362         nip6->ip6_src  = oip6->ip6_src;
363         nip6->ip6_dst  = oip6->ip6_dst;
364
365         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
366                 oip6->ip6_src.s6_addr16[1] = 0;
367         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
368                 oip6->ip6_dst.s6_addr16[1] = 0;
369
370         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
371         icmp6->icmp6_type = type;
372         icmp6->icmp6_code = code;
373         icmp6->icmp6_pptr = htonl((u_int32_t)param);
374
375         /*
376          * icmp6_reflect() is designed to be in the input path.
377          * icmp6_error() can be called from both input and outut path,
378          * and if we are in output path rcvif could contain bogus value.
379          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
380          * information in ip header (nip6).
381          */
382         m->m_pkthdr.rcvif = NULL;
383
384         icmp6stat.icp6s_outhist[type]++;
385         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
386
387         return;
388
389   freeit:
390         /*
391          * If we can't tell wheter or not we can generate ICMP6, free it.
392          */
393         m_freem(m);
394 }
395
396 /*
397  * Process a received ICMP6 message.
398  */
399 int
400 icmp6_input(mp, offp, proto)
401         struct mbuf **mp;
402         int *offp, proto;
403 {
404         struct mbuf *m = *mp, *n;
405         struct ip6_hdr *ip6, *nip6;
406         struct icmp6_hdr *icmp6, *nicmp6;
407         int off = *offp;
408         int icmp6len = m->m_pkthdr.len - *offp;
409         int code, sum, noff;
410
411 #ifndef PULLDOWN_TEST
412         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
413         /* m might change if M_LOOP.  So, call mtod after this */
414 #endif
415
416         /*
417          * Locate icmp6 structure in mbuf, and check
418          * that not corrupted and of at least minimum length
419          */
420
421         ip6 = mtod(m, struct ip6_hdr *);
422         if (icmp6len < sizeof(struct icmp6_hdr)) {
423                 icmp6stat.icp6s_tooshort++;
424                 goto freeit;
425         }
426
427         /*
428          * calculate the checksum
429          */
430 #ifndef PULLDOWN_TEST
431         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
432 #else
433         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
434         if (icmp6 == NULL) {
435                 icmp6stat.icp6s_tooshort++;
436                 return IPPROTO_DONE;
437         }
438 #endif
439         code = icmp6->icmp6_code;
440
441         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
442                 nd6log((LOG_ERR,
443                     "ICMP6 checksum error(%d|%x) %s\n",
444                     icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
445                 icmp6stat.icp6s_checksum++;
446                 goto freeit;
447         }
448
449         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
450                 /*
451                  * Deliver very specific ICMP6 type only.
452                  * This is important to deilver TOOBIG.  Otherwise PMTUD
453                  * will not work.
454                  */
455                 switch (icmp6->icmp6_type) {
456                 case ICMP6_DST_UNREACH:
457                 case ICMP6_PACKET_TOO_BIG:
458                 case ICMP6_TIME_EXCEEDED:
459                         break;
460                 default:
461                         goto freeit;
462                 }
463         }
464
465         icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
466         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
467         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
468                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
469
470         switch (icmp6->icmp6_type) {
471         case ICMP6_DST_UNREACH:
472                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
473                 switch (code) {
474                 case ICMP6_DST_UNREACH_NOROUTE:
475                         code = PRC_UNREACH_NET;
476                         break;
477                 case ICMP6_DST_UNREACH_ADMIN:
478                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
479                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
480                         break;
481                 case ICMP6_DST_UNREACH_ADDR:
482                         code = PRC_HOSTDEAD;
483                         break;
484 #ifdef COMPAT_RFC1885
485                 case ICMP6_DST_UNREACH_NOTNEIGHBOR:
486                         code = PRC_UNREACH_SRCFAIL;
487                         break;
488 #else
489                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
490                         /* I mean "source address was incorrect." */
491                         code = PRC_PARAMPROB;
492                         break;
493 #endif
494                 case ICMP6_DST_UNREACH_NOPORT:
495                         code = PRC_UNREACH_PORT;
496                         break;
497                 default:
498                         goto badcode;
499                 }
500                 goto deliver;
501                 break;
502
503         case ICMP6_PACKET_TOO_BIG:
504                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
505                 if (code != 0)
506                         goto badcode;
507
508                 code = PRC_MSGSIZE;
509
510                 /*
511                  * Updating the path MTU will be done after examining
512                  * intermediate extension headers.
513                  */
514                 goto deliver;
515                 break;
516
517         case ICMP6_TIME_EXCEEDED:
518                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
519                 switch (code) {
520                 case ICMP6_TIME_EXCEED_TRANSIT:
521                 case ICMP6_TIME_EXCEED_REASSEMBLY:
522                         code += PRC_TIMXCEED_INTRANS;
523                         break;
524                 default:
525                         goto badcode;
526                 }
527                 goto deliver;
528                 break;
529
530         case ICMP6_PARAM_PROB:
531                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
532                 switch (code) {
533                 case ICMP6_PARAMPROB_NEXTHEADER:
534                         code = PRC_UNREACH_PROTOCOL;
535                         break;
536                 case ICMP6_PARAMPROB_HEADER:
537                 case ICMP6_PARAMPROB_OPTION:
538                         code = PRC_PARAMPROB;
539                         break;
540                 default:
541                         goto badcode;
542                 }
543                 goto deliver;
544                 break;
545
546         case ICMP6_ECHO_REQUEST:
547                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
548                 if (code != 0)
549                         goto badcode;
550                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
551                         /* Give up remote */
552                         break;
553                 }
554                 if ((n->m_flags & M_EXT) != 0
555                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
556                         struct mbuf *n0 = n;
557                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
558                         int n0len;
559
560                         /*
561                          * Prepare an internal mbuf. m_pullup() doesn't
562                          * always copy the length we specified.
563                          */
564                         if (maxlen >= MCLBYTES) {
565                                 /* Give up remote */
566                                 m_freem(n0);
567                                 break;
568                         }
569                         MGETHDR(n, M_DONTWAIT, n0->m_type);
570                         n0len = n0->m_pkthdr.len;       /* save for use below */
571                         if (n)
572                                 M_MOVE_PKTHDR(n, n0);
573                         if (n && maxlen >= MHLEN) {
574                                 MCLGET(n, M_DONTWAIT);
575                                 if ((n->m_flags & M_EXT) == 0) {
576                                         m_free(n);
577                                         n = NULL;
578                                 }
579                         }
580                         if (n == NULL) {
581                                 /* Give up remote */
582                                 m_freem(n0);
583                                 break;
584                         }
585                         /*
586                          * Copy IPv6 and ICMPv6 only.
587                          */
588                         nip6 = mtod(n, struct ip6_hdr *);
589                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
590                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
591                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
592                         noff = sizeof(struct ip6_hdr);
593                         /* new mbuf contains only ipv6+icmpv6 headers */
594                         n->m_len = noff + sizeof(struct icmp6_hdr);
595                         /*
596                          * Adjust mbuf. ip6_plen will be adjusted in
597                          * ip6_output().
598                          */
599                         m_adj(n0, off + sizeof(struct icmp6_hdr));
600                         /* recalculate complete packet size */
601                         n->m_pkthdr.len = n0len + (noff - off);
602                         n->m_next = n0;
603                 } else {
604                         nip6 = mtod(n, struct ip6_hdr *);
605                         nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
606                         noff = off;
607                 }
608                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
609                 nicmp6->icmp6_code = 0;
610                 if (n) {
611                         icmp6stat.icp6s_reflect++;
612                         icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
613                         icmp6_reflect(n, noff);
614                 }
615                 break;
616
617         case ICMP6_ECHO_REPLY:
618                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
619                 if (code != 0)
620                         goto badcode;
621                 break;
622
623         case MLD_LISTENER_QUERY:
624         case MLD_LISTENER_REPORT:
625                 if (icmp6len < sizeof(struct mld_hdr))
626                         goto badlen;
627                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
628                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
629                 else
630                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
631                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
632                         /* give up local */
633                         mld6_input(m, off);
634                         m = NULL;
635                         goto freeit;
636                 }
637                 mld6_input(n, off);
638                 /* m stays. */
639                 break;
640
641         case MLD_LISTENER_DONE:
642                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
643                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
644                         goto badlen;
645                 break;          /* nothing to be done in kernel */
646
647         case MLD_MTRACE_RESP:
648         case MLD_MTRACE:
649                 /* XXX: these two are experimental.  not officially defind. */
650                 /* XXX: per-interface statistics? */
651                 break;          /* just pass it to applications */
652
653         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
654             {
655                 enum { WRU, FQDN } mode;
656
657                 if (!icmp6_nodeinfo)
658                         break;
659
660                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
661                         mode = WRU;
662                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
663                         mode = FQDN;
664                 else
665                         goto badlen;
666
667 #define hostnamelen     strlen(hostname)
668                 if (mode == FQDN) {
669 #ifndef PULLDOWN_TEST
670                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
671                                          IPPROTO_DONE);
672 #endif
673                         n = m_copy(m, 0, M_COPYALL);
674                         if (n)
675                                 n = ni6_input(n, off);
676                         /* XXX meaningless if n == NULL */
677                         noff = sizeof(struct ip6_hdr);
678                 } else {
679                         u_char *p;
680                         int maxlen, maxhlen;
681
682                         if ((icmp6_nodeinfo & 5) != 5) 
683                                 break;
684
685                         if (code != 0)
686                                 goto badcode;
687                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
688                         if (maxlen >= MCLBYTES) {
689                                 /* Give up remote */
690                                 break;
691                         }
692                         MGETHDR(n, M_DONTWAIT, m->m_type);
693                         if (n && maxlen > MHLEN) {
694                                 MCLGET(n, M_DONTWAIT);
695                                 if ((n->m_flags & M_EXT) == 0) {
696                                         m_free(n);
697                                         n = NULL;
698                                 }
699                         }
700                         if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
701                                 /*
702                                  * Previous code did a blind M_COPY_PKTHDR
703                                  * and said "just for rcvif".  If true, then
704                                  * we could tolerate the dup failing (due to
705                                  * the deep copy of the tag chain).  For now
706                                  * be conservative and just fail.
707                                  */
708                                 m_free(n);
709                                 n = NULL;
710                         }
711                         if (n == NULL) {
712                                 /* Give up remote */
713                                 break;
714                         }
715                         n->m_pkthdr.rcvif = NULL;
716                         n->m_len = 0;
717                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
718                         if (maxhlen > hostnamelen)
719                                 maxhlen = hostnamelen;
720                         /*
721                          * Copy IPv6 and ICMPv6 only.
722                          */
723                         nip6 = mtod(n, struct ip6_hdr *);
724                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
725                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
726                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
727                         p = (u_char *)(nicmp6 + 1);
728                         bzero(p, 4);
729                         bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
730                         noff = sizeof(struct ip6_hdr);
731                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
732                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
733                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
734                         nicmp6->icmp6_code = 0;
735                 }
736 #undef hostnamelen
737                 if (n) {
738                         icmp6stat.icp6s_reflect++;
739                         icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
740                         icmp6_reflect(n, noff);
741                 }
742                 break;
743             }
744
745         case ICMP6_WRUREPLY:
746                 if (code != 0)
747                         goto badcode;
748                 break;
749
750         case ND_ROUTER_SOLICIT:
751                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
752                 if (code != 0)
753                         goto badcode;
754                 if (icmp6len < sizeof(struct nd_router_solicit))
755                         goto badlen;
756                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
757                         /* give up local */
758                         nd6_rs_input(m, off, icmp6len);
759                         m = NULL;
760                         goto freeit;
761                 }
762                 nd6_rs_input(n, off, icmp6len);
763                 /* m stays. */
764                 break;
765
766         case ND_ROUTER_ADVERT:
767                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
768                 if (code != 0)
769                         goto badcode;
770                 if (icmp6len < sizeof(struct nd_router_advert))
771                         goto badlen;
772                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
773                         /* give up local */
774                         nd6_ra_input(m, off, icmp6len);
775                         m = NULL;
776                         goto freeit;
777                 }
778                 nd6_ra_input(n, off, icmp6len);
779                 /* m stays. */
780                 break;
781
782         case ND_NEIGHBOR_SOLICIT:
783                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
784                 if (code != 0)
785                         goto badcode;
786                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
787                         goto badlen;
788                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
789                         /* give up local */
790                         nd6_ns_input(m, off, icmp6len);
791                         m = NULL;
792                         goto freeit;
793                 }
794                 nd6_ns_input(n, off, icmp6len);
795                 /* m stays. */
796                 break;
797
798         case ND_NEIGHBOR_ADVERT:
799                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
800                 if (code != 0)
801                         goto badcode;
802                 if (icmp6len < sizeof(struct nd_neighbor_advert))
803                         goto badlen;
804                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
805                         /* give up local */
806                         nd6_na_input(m, off, icmp6len);
807                         m = NULL;
808                         goto freeit;
809                 }
810                 nd6_na_input(n, off, icmp6len);
811                 /* m stays. */
812                 break;
813
814         case ND_REDIRECT:
815                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
816                 if (code != 0)
817                         goto badcode;
818                 if (icmp6len < sizeof(struct nd_redirect))
819                         goto badlen;
820                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
821                         /* give up local */
822                         icmp6_redirect_input(m, off);
823                         m = NULL;
824                         goto freeit;
825                 }
826                 icmp6_redirect_input(n, off);
827                 /* m stays. */
828                 break;
829
830         case ICMP6_ROUTER_RENUMBERING:
831                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
832                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
833                         goto badcode;
834                 if (icmp6len < sizeof(struct icmp6_router_renum))
835                         goto badlen;
836                 break;
837
838         default:
839                 nd6log((LOG_DEBUG,
840                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
841                     icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
842                     ip6_sprintf(&ip6->ip6_dst),
843                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
844                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
845                         /* ICMPv6 error: MUST deliver it by spec... */
846                         code = PRC_NCMDS;
847                         /* deliver */
848                 } else {
849                         /* ICMPv6 informational: MUST not deliver */
850                         break;
851                 }
852         deliver:
853                 if (icmp6_notify_error(m, off, icmp6len, code)) {
854                         /* In this case, m should've been freed. */
855                         return(IPPROTO_DONE);
856                 }
857                 break;
858
859         badcode:
860                 icmp6stat.icp6s_badcode++;
861                 break;
862
863         badlen:
864                 icmp6stat.icp6s_badlen++;
865                 break;
866         }
867
868         /* deliver the packet to appropriate sockets */
869         icmp6_rip6_input(&m, *offp);
870
871         return IPPROTO_DONE;
872
873  freeit:
874         m_freem(m);
875         return IPPROTO_DONE;
876 }
877
878 static int
879 icmp6_notify_error(m, off, icmp6len, code)
880         struct mbuf *m;
881         int off, icmp6len;
882 {
883         struct icmp6_hdr *icmp6;
884         struct ip6_hdr *eip6;
885         u_int32_t notifymtu;
886         struct sockaddr_in6 icmp6src, icmp6dst;
887
888         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
889                 icmp6stat.icp6s_tooshort++;
890                 goto freeit;
891         }
892 #ifndef PULLDOWN_TEST
893         IP6_EXTHDR_CHECK(m, off,
894                          sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
895                          -1);
896         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
897 #else
898         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
899                        sizeof(*icmp6) + sizeof(struct ip6_hdr));
900         if (icmp6 == NULL) {
901                 icmp6stat.icp6s_tooshort++;
902                 return(-1);
903         }
904 #endif
905         eip6 = (struct ip6_hdr *)(icmp6 + 1);
906
907         /* Detect the upper level protocol */
908         {
909                 void (*ctlfunc) (int, struct sockaddr *, void *);
910                 u_int8_t nxt = eip6->ip6_nxt;
911                 int eoff = off + sizeof(struct icmp6_hdr) +
912                         sizeof(struct ip6_hdr);
913                 struct ip6ctlparam ip6cp;
914                 struct in6_addr *finaldst = NULL;
915                 int icmp6type = icmp6->icmp6_type;
916                 struct ip6_frag *fh;
917                 struct ip6_rthdr *rth;
918                 struct ip6_rthdr0 *rth0;
919                 int rthlen;
920
921                 while (1) { /* XXX: should avoid infinite loop explicitly? */
922                         struct ip6_ext *eh;
923
924                         switch (nxt) {
925                         case IPPROTO_HOPOPTS:
926                         case IPPROTO_DSTOPTS:
927                         case IPPROTO_AH:
928 #ifndef PULLDOWN_TEST
929                                 IP6_EXTHDR_CHECK(m, 0, eoff +
930                                                  sizeof(struct ip6_ext),
931                                                  -1);
932                                 eh = (struct ip6_ext *)(mtod(m, caddr_t)
933                                                         + eoff);
934 #else
935                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
936                                                eoff, sizeof(*eh));
937                                 if (eh == NULL) {
938                                         icmp6stat.icp6s_tooshort++;
939                                         return(-1);
940                                 }
941 #endif
942                                 
943                                 if (nxt == IPPROTO_AH)
944                                         eoff += (eh->ip6e_len + 2) << 2;
945                                 else
946                                         eoff += (eh->ip6e_len + 1) << 3;
947                                 nxt = eh->ip6e_nxt;
948                                 break;
949                         case IPPROTO_ROUTING:
950                                 /*
951                                  * When the erroneous packet contains a
952                                  * routing header, we should examine the
953                                  * header to determine the final destination.
954                                  * Otherwise, we can't properly update
955                                  * information that depends on the final
956                                  * destination (e.g. path MTU).
957                                  */
958 #ifndef PULLDOWN_TEST
959                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
960                                                  -1);
961                                 rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
962                                                            + eoff);
963 #else
964                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
965                                                eoff, sizeof(*rth));
966                                 if (rth == NULL) {
967                                         icmp6stat.icp6s_tooshort++;
968                                         return(-1);
969                                 }
970 #endif
971                                 rthlen = (rth->ip6r_len + 1) << 3;
972                                 /*
973                                  * XXX: currently there is no
974                                  * officially defined type other
975                                  * than type-0.
976                                  * Note that if the segment left field
977                                  * is 0, all intermediate hops must
978                                  * have been passed.
979                                  */
980                                 if (rth->ip6r_segleft &&
981                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
982                                         int hops;
983
984 #ifndef PULLDOWN_TEST
985                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
986                                                          -1);
987                                         rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
988 #else
989                                         IP6_EXTHDR_GET(rth0,
990                                                        struct ip6_rthdr0 *, m,
991                                                        eoff, rthlen);
992                                         if (rth0 == NULL) {
993                                                 icmp6stat.icp6s_tooshort++;
994                                                 return(-1);
995                                         }
996 #endif
997                                         /* just ignore a bogus header */
998                                         if ((rth0->ip6r0_len % 2) == 0 &&
999                                             (hops = rth0->ip6r0_len/2))
1000                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1001                                 }
1002                                 eoff += rthlen;
1003                                 nxt = rth->ip6r_nxt;
1004                                 break;
1005                         case IPPROTO_FRAGMENT:
1006 #ifndef PULLDOWN_TEST
1007                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1008                                                  sizeof(struct ip6_frag),
1009                                                  -1);
1010                                 fh = (struct ip6_frag *)(mtod(m, caddr_t)
1011                                                          + eoff);
1012 #else
1013                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1014                                                eoff, sizeof(*fh));
1015                                 if (fh == NULL) {
1016                                         icmp6stat.icp6s_tooshort++;
1017                                         return(-1);
1018                                 }
1019 #endif
1020                                 /*
1021                                  * Data after a fragment header is meaningless
1022                                  * unless it is the first fragment, but
1023                                  * we'll go to the notify label for path MTU
1024                                  * discovery.
1025                                  */
1026                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1027                                         goto notify;
1028
1029                                 eoff += sizeof(struct ip6_frag);
1030                                 nxt = fh->ip6f_nxt;
1031                                 break;
1032                         default:
1033                                 /*
1034                                  * This case includes ESP and the No Next
1035                                  * Header.  In such cases going to the notify
1036                                  * label does not have any meaning
1037                                  * (i.e. ctlfunc will be NULL), but we go
1038                                  * anyway since we might have to update
1039                                  * path MTU information.
1040                                  */
1041                                 goto notify;
1042                         }
1043                 }
1044           notify:
1045 #ifndef PULLDOWN_TEST
1046                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1047 #else
1048                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1049                                sizeof(*icmp6) + sizeof(struct ip6_hdr));
1050                 if (icmp6 == NULL) {
1051                         icmp6stat.icp6s_tooshort++;
1052                         return(-1);
1053                 }
1054 #endif
1055
1056                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1057                 bzero(&icmp6dst, sizeof(icmp6dst));
1058                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1059                 icmp6dst.sin6_family = AF_INET6;
1060                 if (finaldst == NULL)
1061                         icmp6dst.sin6_addr = eip6->ip6_dst;
1062                 else
1063                         icmp6dst.sin6_addr = *finaldst;
1064                 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1065                                                           &icmp6dst.sin6_addr);
1066 #ifndef SCOPEDROUTING
1067                 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1068                                    NULL, NULL)) {
1069                         /* should be impossbile */
1070                         nd6log((LOG_DEBUG,
1071                             "icmp6_notify_error: in6_embedscope failed\n"));
1072                         goto freeit;
1073                 }
1074 #endif
1075
1076                 /*
1077                  * retrieve parameters from the inner IPv6 header, and convert
1078                  * them into sockaddr structures.
1079                  */
1080                 bzero(&icmp6src, sizeof(icmp6src));
1081                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1082                 icmp6src.sin6_family = AF_INET6;
1083                 icmp6src.sin6_addr = eip6->ip6_src;
1084                 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1085                                                           &icmp6src.sin6_addr);
1086 #ifndef SCOPEDROUTING
1087                 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1088                                    NULL, NULL)) {
1089                         /* should be impossbile */
1090                         nd6log((LOG_DEBUG,
1091                             "icmp6_notify_error: in6_embedscope failed\n"));
1092                         goto freeit;
1093                 }
1094 #endif
1095                 icmp6src.sin6_flowinfo =
1096                         (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1097
1098                 if (finaldst == NULL)
1099                         finaldst = &eip6->ip6_dst;
1100                 ip6cp.ip6c_m = m;
1101                 ip6cp.ip6c_icmp6 = icmp6;
1102                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1103                 ip6cp.ip6c_off = eoff;
1104                 ip6cp.ip6c_finaldst = finaldst;
1105                 ip6cp.ip6c_src = &icmp6src;
1106                 ip6cp.ip6c_nxt = nxt;
1107
1108                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1109                         notifymtu = ntohl(icmp6->icmp6_mtu);
1110                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1111                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1112                 }
1113
1114                 ctlfunc = (void (*) (int, struct sockaddr *, void *))
1115                         (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1116                 if (ctlfunc) {
1117                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1118                                           &ip6cp);
1119                 }
1120         }
1121         return(0);
1122
1123   freeit:
1124         m_freem(m);
1125         return(-1);
1126 }
1127
1128 void
1129 icmp6_mtudisc_update(ip6cp, validated)
1130         struct ip6ctlparam *ip6cp;
1131         int validated;
1132 {
1133         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1134         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1135         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1136         u_int mtu = ntohl(icmp6->icmp6_mtu);
1137         struct rtentry *rt = NULL;
1138         struct sockaddr_in6 sin6;
1139
1140         if (!validated)
1141                 return;
1142
1143         bzero(&sin6, sizeof(sin6));
1144         sin6.sin6_family = PF_INET6;
1145         sin6.sin6_len = sizeof(struct sockaddr_in6);
1146         sin6.sin6_addr = *dst;
1147         /* XXX normally, this won't happen */
1148         if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1149                 sin6.sin6_addr.s6_addr16[1] =
1150                     htons(m->m_pkthdr.rcvif->if_index);
1151         }
1152         /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1153         rt = rtalloc1((struct sockaddr *)&sin6, 0,
1154                       RTF_CLONING | RTF_PRCLONING);
1155
1156         if (rt && (rt->rt_flags & RTF_HOST)
1157             && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1158                 if (mtu < IPV6_MMTU) {
1159                                 /* xxx */
1160                         rt->rt_rmx.rmx_locks |= RTV_MTU;
1161                 } else if (mtu < rt->rt_ifp->if_mtu &&
1162                            rt->rt_rmx.rmx_mtu > mtu) {
1163                         icmp6stat.icp6s_pmtuchg++;
1164                         rt->rt_rmx.rmx_mtu = mtu;
1165                 }
1166         }
1167         if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1168                 RTFREE(rt);
1169         }
1170 }
1171
1172 /*
1173  * Process a Node Information Query packet, based on
1174  * draft-ietf-ipngwg-icmp-name-lookups-07.
1175  * 
1176  * Spec incompatibilities:
1177  * - IPv6 Subject address handling
1178  * - IPv4 Subject address handling support missing
1179  * - Proxy reply (answer even if it's not for me)
1180  * - joins NI group address at in6_ifattach() time only, does not cope
1181  *   with hostname changes by sethostname(3)
1182  */
1183 #define hostnamelen     strlen(hostname)
1184 static struct mbuf *
1185 ni6_input(m, off)
1186         struct mbuf *m;
1187         int off;
1188 {
1189         struct icmp6_nodeinfo *ni6, *nni6;
1190         struct mbuf *n = NULL;
1191         u_int16_t qtype;
1192         int subjlen;
1193         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1194         struct ni_reply_fqdn *fqdn;
1195         int addrs;              /* for NI_QTYPE_NODEADDR */
1196         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1197         struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1198         struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1199         struct ip6_hdr *ip6;
1200         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1201         char *subj = NULL;
1202         struct in6_ifaddr *ia6 = NULL;
1203
1204         ip6 = mtod(m, struct ip6_hdr *);
1205 #ifndef PULLDOWN_TEST
1206         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1207 #else
1208         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1209         if (ni6 == NULL) {
1210                 /* m is already reclaimed */
1211                 return NULL;
1212         }
1213 #endif
1214
1215         /*
1216          * Validate IPv6 destination address.
1217          *
1218          * The Responder must discard the Query without further processing
1219          * unless it is one of the Responder's unicast or anycast addresses, or
1220          * a link-local scope multicast address which the Responder has joined.
1221          * [icmp-name-lookups-07, Section 4.]
1222          */
1223         bzero(&sin6, sizeof(sin6));
1224         sin6.sin6_family = AF_INET6;
1225         sin6.sin6_len = sizeof(struct sockaddr_in6);
1226         bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1227         /* XXX scopeid */
1228         if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1229                 /* unicast/anycast, fine */
1230                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1231                     (icmp6_nodeinfo & 4) == 0) {
1232                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1233                                 "a temporary address in %s:%d",
1234                                __FILE__, __LINE__));
1235                         goto bad;
1236                 }
1237         } else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1238                 ; /* link-local multicast, fine */
1239         else
1240                 goto bad;
1241
1242         /* validate query Subject field. */
1243         qtype = ntohs(ni6->ni_qtype);
1244         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1245         switch (qtype) {
1246         case NI_QTYPE_NOOP:
1247         case NI_QTYPE_SUPTYPES:
1248                 /* 07 draft */
1249                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1250                         break;
1251                 /* FALLTHROUGH */
1252         case NI_QTYPE_FQDN:
1253         case NI_QTYPE_NODEADDR:
1254                 switch (ni6->ni_code) {
1255                 case ICMP6_NI_SUBJ_IPV6:
1256 #if ICMP6_NI_SUBJ_IPV6 != 0
1257                 case 0:
1258 #endif
1259                         /*
1260                          * backward compatibility - try to accept 03 draft
1261                          * format, where no Subject is present.
1262                          */
1263                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1264                             subjlen == 0) {
1265                                 oldfqdn++;
1266                                 break;
1267                         }
1268 #if ICMP6_NI_SUBJ_IPV6 != 0
1269                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1270                                 goto bad;
1271 #endif
1272
1273                         if (subjlen != sizeof(sin6.sin6_addr))
1274                                 goto bad;
1275
1276                         /*
1277                          * Validate Subject address.
1278                          *
1279                          * Not sure what exactly "address belongs to the node"
1280                          * means in the spec, is it just unicast, or what?
1281                          *
1282                          * At this moment we consider Subject address as
1283                          * "belong to the node" if the Subject address equals
1284                          * to the IPv6 destination address; validation for
1285                          * IPv6 destination address should have done enough
1286                          * check for us.
1287                          *
1288                          * We do not do proxy at this moment.
1289                          */
1290                         /* m_pulldown instead of copy? */
1291                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1292                             subjlen, (caddr_t)&sin6.sin6_addr);
1293                         sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1294                                                               &sin6.sin6_addr);
1295 #ifndef SCOPEDROUTING
1296                         in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1297 #endif
1298                         bzero(&sin6_d, sizeof(sin6_d));
1299                         sin6_d.sin6_family = AF_INET6; /* not used, actually */
1300                         sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1301                         sin6_d.sin6_addr = ip6->ip6_dst;
1302                         sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1303                                                                 &ip6->ip6_dst);
1304 #ifndef SCOPEDROUTING
1305                         in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1306 #endif
1307                         subj = (char *)&sin6;
1308                         if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1309                                 break;
1310
1311                         /*
1312                          * XXX if we are to allow other cases, we should really
1313                          * be careful about scope here.
1314                          * basically, we should disallow queries toward IPv6
1315                          * destination X with subject Y, if scope(X) > scope(Y).
1316                          * if we allow scope(X) > scope(Y), it will result in
1317                          * information leakage across scope boundary.
1318                          */
1319                         goto bad;
1320
1321                 case ICMP6_NI_SUBJ_FQDN:
1322                         /*
1323                          * Validate Subject name with gethostname(3).
1324                          *
1325                          * The behavior may need some debate, since:
1326                          * - we are not sure if the node has FQDN as
1327                          *   hostname (returned by gethostname(3)).
1328                          * - the code does wildcard match for truncated names.
1329                          *   however, we are not sure if we want to perform
1330                          *   wildcard match, if gethostname(3) side has
1331                          *   truncated hostname.
1332                          */
1333                         n = ni6_nametodns(hostname, hostnamelen, 0);
1334                         if (!n || n->m_next || n->m_len == 0)
1335                                 goto bad;
1336                         IP6_EXTHDR_GET(subj, char *, m,
1337                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1338                         if (subj == NULL)
1339                                 goto bad;
1340                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1341                                         n->m_len)) {
1342                                 goto bad;
1343                         }
1344                         m_freem(n);
1345                         n = NULL;
1346                         break;
1347
1348                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1349                 default:
1350                         goto bad;
1351                 }
1352                 break;
1353         }
1354
1355         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1356         switch (qtype) {
1357         case NI_QTYPE_FQDN:
1358                 if ((icmp6_nodeinfo & 1) == 0)
1359                         goto bad;
1360                 break;
1361         case NI_QTYPE_NODEADDR:
1362                 if ((icmp6_nodeinfo & 2) == 0)
1363                         goto bad;
1364                 break;
1365         }
1366
1367         /* guess reply length */
1368         switch (qtype) {
1369         case NI_QTYPE_NOOP:
1370                 break;          /* no reply data */
1371         case NI_QTYPE_SUPTYPES:
1372                 replylen += sizeof(u_int32_t);
1373                 break;
1374         case NI_QTYPE_FQDN:
1375                 /* XXX will append an mbuf */
1376                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1377                 break;
1378         case NI_QTYPE_NODEADDR:
1379                 addrs = ni6_addrs(ni6, m, &ifp, subj);
1380                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1381                                           sizeof(u_int32_t))) > MCLBYTES)
1382                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1383                 break;
1384         default:
1385                 /*
1386                  * XXX: We must return a reply with the ICMP6 code
1387                  * `unknown Qtype' in this case. However we regard the case
1388                  * as an FQDN query for backward compatibility.
1389                  * Older versions set a random value to this field,
1390                  * so it rarely varies in the defined qtypes.
1391                  * But the mechanism is not reliable...
1392                  * maybe we should obsolete older versions.
1393                  */
1394                 qtype = NI_QTYPE_FQDN;
1395                 /* XXX will append an mbuf */
1396                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1397                 oldfqdn++;
1398                 break;
1399         }
1400
1401         /* allocate an mbuf to reply. */
1402         MGETHDR(n, M_DONTWAIT, m->m_type);
1403         if (n == NULL) {
1404                 m_freem(m);
1405                 return(NULL);
1406         }
1407         M_MOVE_PKTHDR(n, m); /* just for recvif */
1408         if (replylen > MHLEN) {
1409                 if (replylen > MCLBYTES) {
1410                         /*
1411                          * XXX: should we try to allocate more? But MCLBYTES
1412                          * is probably much larger than IPV6_MMTU...
1413                          */
1414                         goto bad;
1415                 }
1416                 MCLGET(n, M_DONTWAIT);
1417                 if ((n->m_flags & M_EXT) == 0) {
1418                         goto bad;
1419                 }
1420         }
1421         n->m_pkthdr.len = n->m_len = replylen;
1422
1423         /* copy mbuf header and IPv6 + Node Information base headers */
1424         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1425         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1426         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1427
1428         /* qtype dependent procedure */
1429         switch (qtype) {
1430         case NI_QTYPE_NOOP:
1431                 nni6->ni_code = ICMP6_NI_SUCCESS;
1432                 nni6->ni_flags = 0;
1433                 break;
1434         case NI_QTYPE_SUPTYPES:
1435         {
1436                 u_int32_t v;
1437                 nni6->ni_code = ICMP6_NI_SUCCESS;
1438                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1439                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1440                 v = (u_int32_t)htonl(0x0000000f);
1441                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1442                 break;
1443         }
1444         case NI_QTYPE_FQDN:
1445                 nni6->ni_code = ICMP6_NI_SUCCESS;
1446                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1447                                                 sizeof(struct ip6_hdr) +
1448                                                 sizeof(struct icmp6_nodeinfo));
1449                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1450                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1451                 /*
1452                  * XXX do we really have FQDN in variable "hostname"?
1453                  */
1454                 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1455                 if (n->m_next == NULL)
1456                         goto bad;
1457                 /* XXX we assume that n->m_next is not a chain */
1458                 if (n->m_next->m_next != NULL)
1459                         goto bad;
1460                 n->m_pkthdr.len += n->m_next->m_len;
1461                 break;
1462         case NI_QTYPE_NODEADDR:
1463         {
1464                 int lenlim, copied;
1465
1466                 nni6->ni_code = ICMP6_NI_SUCCESS;
1467                 n->m_pkthdr.len = n->m_len =
1468                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1469                 lenlim = M_TRAILINGSPACE(n);
1470                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1471                 /* XXX: reset mbuf length */
1472                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1473                         sizeof(struct icmp6_nodeinfo) + copied;
1474                 break;
1475         }
1476         default:
1477                 break;          /* XXX impossible! */
1478         }
1479
1480         nni6->ni_type = ICMP6_NI_REPLY;
1481         m_freem(m);
1482         return(n);
1483
1484   bad:
1485         m_freem(m);
1486         if (n)
1487                 m_freem(n);
1488         return(NULL);
1489 }
1490 #undef hostnamelen
1491
1492 /*
1493  * make a mbuf with DNS-encoded string.  no compression support.
1494  *
1495  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1496  * treated as truncated name (two \0 at the end).  this is a wild guess.
1497  */
1498 static struct mbuf *
1499 ni6_nametodns(name, namelen, old)
1500         const char *name;
1501         int namelen;
1502         int old;        /* return pascal string if non-zero */
1503 {
1504         struct mbuf *m;
1505         char *cp, *ep;
1506         const char *p, *q;
1507         int i, len, nterm;
1508
1509         if (old)
1510                 len = namelen + 1;
1511         else
1512                 len = MCLBYTES;
1513
1514         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1515         MGET(m, M_DONTWAIT, MT_DATA);
1516         if (m && len > MLEN) {
1517                 MCLGET(m, M_DONTWAIT);
1518                 if ((m->m_flags & M_EXT) == 0)
1519                         goto fail;
1520         }
1521         if (!m)
1522                 goto fail;
1523         m->m_next = NULL;
1524
1525         if (old) {
1526                 m->m_len = len;
1527                 *mtod(m, char *) = namelen;
1528                 bcopy(name, mtod(m, char *) + 1, namelen);
1529                 return m;
1530         } else {
1531                 m->m_len = 0;
1532                 cp = mtod(m, char *);
1533                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1534
1535                 /* if not certain about my name, return empty buffer */
1536                 if (namelen == 0)
1537                         return m;
1538
1539                 /*
1540                  * guess if it looks like shortened hostname, or FQDN.
1541                  * shortened hostname needs two trailing "\0".
1542                  */
1543                 i = 0;
1544                 for (p = name; p < name + namelen; p++) {
1545                         if (*p && *p == '.')
1546                                 i++;
1547                 }
1548                 if (i < 2)
1549                         nterm = 2;
1550                 else
1551                         nterm = 1;
1552
1553                 p = name;
1554                 while (cp < ep && p < name + namelen) {
1555                         i = 0;
1556                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1557                                 i++;
1558                         /* result does not fit into mbuf */
1559                         if (cp + i + 1 >= ep)
1560                                 goto fail;
1561                         /*
1562                          * DNS label length restriction, RFC1035 page 8.
1563                          * "i == 0" case is included here to avoid returning
1564                          * 0-length label on "foo..bar".
1565                          */
1566                         if (i <= 0 || i >= 64)
1567                                 goto fail;
1568                         *cp++ = i;
1569                         bcopy(p, cp, i);
1570                         cp += i;
1571                         p = q;
1572                         if (p < name + namelen && *p == '.')
1573                                 p++;
1574                 }
1575                 /* termination */
1576                 if (cp + nterm >= ep)
1577                         goto fail;
1578                 while (nterm-- > 0)
1579                         *cp++ = '\0';
1580                 m->m_len = cp - mtod(m, char *);
1581                 return m;
1582         }
1583
1584         panic("should not reach here");
1585         /* NOTREACHED */
1586
1587  fail:
1588         if (m)
1589                 m_freem(m);
1590         return NULL;
1591 }
1592
1593 /*
1594  * check if two DNS-encoded string matches.  takes care of truncated
1595  * form (with \0\0 at the end).  no compression support.
1596  * XXX upper/lowercase match (see RFC2065)
1597  */
1598 static int
1599 ni6_dnsmatch(a, alen, b, blen)
1600         const char *a;
1601         int alen;
1602         const char *b;
1603         int blen;
1604 {
1605         const char *a0, *b0;
1606         int l;
1607
1608         /* simplest case - need validation? */
1609         if (alen == blen && bcmp(a, b, alen) == 0)
1610                 return 1;
1611
1612         a0 = a;
1613         b0 = b;
1614
1615         /* termination is mandatory */
1616         if (alen < 2 || blen < 2)
1617                 return 0;
1618         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1619                 return 0;
1620         alen--;
1621         blen--;
1622
1623         while (a - a0 < alen && b - b0 < blen) {
1624                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1625                         return 0;
1626
1627                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1628                         return 0;
1629                 /* we don't support compression yet */
1630                 if (a[0] >= 64 || b[0] >= 64)
1631                         return 0;
1632
1633                 /* truncated case */
1634                 if (a[0] == 0 && a - a0 == alen - 1)
1635                         return 1;
1636                 if (b[0] == 0 && b - b0 == blen - 1)
1637                         return 1;
1638                 if (a[0] == 0 || b[0] == 0)
1639                         return 0;
1640
1641                 if (a[0] != b[0])
1642                         return 0;
1643                 l = a[0];
1644                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1645                         return 0;
1646                 if (bcmp(a + 1, b + 1, l) != 0)
1647                         return 0;
1648
1649                 a += 1 + l;
1650                 b += 1 + l;
1651         }
1652
1653         if (a - a0 == alen && b - b0 == blen)
1654                 return 1;
1655         else
1656                 return 0;
1657 }
1658
1659 /*
1660  * calculate the number of addresses to be returned in the node info reply.
1661  */
1662 static int
1663 ni6_addrs(ni6, m, ifpp, subj)
1664         struct icmp6_nodeinfo *ni6;
1665         struct mbuf *m;
1666         struct ifnet **ifpp;
1667         char *subj;
1668 {
1669         struct ifnet *ifp;
1670         struct in6_ifaddr *ifa6;
1671         struct ifaddr *ifa;
1672         struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1673         int addrs = 0, addrsofif, iffound = 0;
1674         int niflags = ni6->ni_flags;
1675
1676         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1677                 switch (ni6->ni_code) {
1678                 case ICMP6_NI_SUBJ_IPV6:
1679                         if (subj == NULL) /* must be impossible... */
1680                                 return(0);
1681                         subj_ip6 = (struct sockaddr_in6 *)subj;
1682                         break;
1683                 default:
1684                         /*
1685                          * XXX: we only support IPv6 subject address for
1686                          * this Qtype.
1687                          */
1688                         return(0);
1689                 }
1690         }
1691
1692         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1693         {
1694                 addrsofif = 0;
1695                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1696                 {
1697                         if (ifa->ifa_addr->sa_family != AF_INET6)
1698                                 continue;
1699                         ifa6 = (struct in6_ifaddr *)ifa;
1700
1701                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1702                             IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1703                                                &ifa6->ia_addr.sin6_addr))
1704                                 iffound = 1;
1705
1706                         /*
1707                          * IPv4-mapped addresses can only be returned by a
1708                          * Node Information proxy, since they represent
1709                          * addresses of IPv4-only nodes, which perforce do
1710                          * not implement this protocol.
1711                          * [icmp-name-lookups-07, Section 5.4]
1712                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1713                          * this function at this moment.
1714                          */
1715
1716                         /* What do we have to do about ::1? */
1717                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1718                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1719                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1720                                         continue;
1721                                 break;
1722                         case IPV6_ADDR_SCOPE_SITELOCAL:
1723                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1724                                         continue;
1725                                 break;
1726                         case IPV6_ADDR_SCOPE_GLOBAL:
1727                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1728                                         continue;
1729                                 break;
1730                         default:
1731                                 continue;
1732                         }
1733
1734                         /*
1735                          * check if anycast is okay.
1736                          * XXX: just experimental.  not in the spec.
1737                          */
1738                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1739                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1740                                 continue; /* we need only unicast addresses */
1741                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1742                             (icmp6_nodeinfo & 4) == 0) {
1743                                 continue;
1744                         }
1745                         addrsofif++; /* count the address */
1746                 }
1747                 if (iffound) {
1748                         *ifpp = ifp;
1749                         return(addrsofif);
1750                 }
1751
1752                 addrs += addrsofif;
1753         }
1754
1755         return(addrs);
1756 }
1757
1758 static int
1759 ni6_store_addrs(ni6, nni6, ifp0, resid)
1760         struct icmp6_nodeinfo *ni6, *nni6;
1761         struct ifnet *ifp0;
1762         int resid;
1763 {
1764         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1765         struct in6_ifaddr *ifa6;
1766         struct ifaddr *ifa;
1767         struct ifnet *ifp_dep = NULL;
1768         int copied = 0, allow_deprecated = 0;
1769         u_char *cp = (u_char *)(nni6 + 1);
1770         int niflags = ni6->ni_flags;
1771         u_int32_t ltime;
1772
1773         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1774                 return(0);      /* needless to copy */
1775
1776   again:
1777
1778         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1779         {
1780                 for (ifa = ifp->if_addrlist.tqh_first; ifa;
1781                      ifa = ifa->ifa_list.tqe_next)
1782                 {
1783                         if (ifa->ifa_addr->sa_family != AF_INET6)
1784                                 continue;
1785                         ifa6 = (struct in6_ifaddr *)ifa;
1786
1787                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1788                             allow_deprecated == 0) {
1789                                 /*
1790                                  * prefererred address should be put before
1791                                  * deprecated addresses.
1792                                  */
1793
1794                                 /* record the interface for later search */
1795                                 if (ifp_dep == NULL)
1796                                         ifp_dep = ifp;
1797
1798                                 continue;
1799                         }
1800                         else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1801                                  allow_deprecated != 0)
1802                                 continue; /* we now collect deprecated addrs */
1803
1804                         /* What do we have to do about ::1? */
1805                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1806                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1807                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1808                                         continue;
1809                                 break;
1810                         case IPV6_ADDR_SCOPE_SITELOCAL:
1811                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1812                                         continue;
1813                                 break;
1814                         case IPV6_ADDR_SCOPE_GLOBAL:
1815                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1816                                         continue;
1817                                 break;
1818                         default:
1819                                 continue;
1820                         }
1821
1822                         /*
1823                          * check if anycast is okay.
1824                          * XXX: just experimental. not in the spec.
1825                          */
1826                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1827                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1828                                 continue;
1829                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1830                             (icmp6_nodeinfo & 4) == 0) {
1831                                 continue;
1832                         }
1833
1834                         /* now we can copy the address */
1835                         if (resid < sizeof(struct in6_addr) +
1836                             sizeof(u_int32_t)) {
1837                                 /*
1838                                  * We give up much more copy.
1839                                  * Set the truncate flag and return.
1840                                  */
1841                                 nni6->ni_flags |=
1842                                         NI_NODEADDR_FLAG_TRUNCATE;
1843                                 return(copied);
1844                         }
1845
1846                         /*
1847                          * Set the TTL of the address.
1848                          * The TTL value should be one of the following
1849                          * according to the specification:
1850                          *
1851                          * 1. The remaining lifetime of a DHCP lease on the
1852                          *    address, or
1853                          * 2. The remaining Valid Lifetime of a prefix from
1854                          *    which the address was derived through Stateless
1855                          *    Autoconfiguration.
1856                          *
1857                          * Note that we currently do not support stateful
1858                          * address configuration by DHCPv6, so the former
1859                          * case can't happen.
1860                          */
1861                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1862                                 ltime = ND6_INFINITE_LIFETIME;
1863                         else {
1864                                 if (ifa6->ia6_lifetime.ia6t_expire >
1865                                     time_second)
1866                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1867                                 else
1868                                         ltime = 0;
1869                         }
1870                         
1871                         bcopy(&ltime, cp, sizeof(u_int32_t));
1872                         cp += sizeof(u_int32_t);
1873
1874                         /* copy the address itself */
1875                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1876                               sizeof(struct in6_addr));
1877                         /* XXX: KAME link-local hack; remove ifindex */
1878                         if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1879                                 ((struct in6_addr *)cp)->s6_addr16[1] = 0;
1880                         cp += sizeof(struct in6_addr);
1881                         
1882                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1883                         copied += (sizeof(struct in6_addr) +
1884                                    sizeof(u_int32_t));
1885                 }
1886                 if (ifp0)       /* we need search only on the specified IF */
1887                         break;
1888         }
1889
1890         if (allow_deprecated == 0 && ifp_dep != NULL) {
1891                 ifp = ifp_dep;
1892                 allow_deprecated = 1;
1893
1894                 goto again;
1895         }
1896
1897         return(copied);
1898 }
1899
1900 /*
1901  * XXX almost dup'ed code with rip6_input.
1902  */
1903 static int
1904 icmp6_rip6_input(mp, off)
1905         struct  mbuf **mp;
1906         int     off;
1907 {
1908         struct mbuf *m = *mp;
1909         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1910         struct in6pcb *in6p;
1911         struct in6pcb *last = NULL;
1912         struct sockaddr_in6 rip6src;
1913         struct icmp6_hdr *icmp6;
1914         struct mbuf *opts = NULL;
1915
1916 #ifndef PULLDOWN_TEST
1917         /* this is assumed to be safe. */
1918         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1919 #else
1920         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1921         if (icmp6 == NULL) {
1922                 /* m is already reclaimed */
1923                 return IPPROTO_DONE;
1924         }
1925 #endif
1926
1927         bzero(&rip6src, sizeof(rip6src));
1928         rip6src.sin6_len = sizeof(struct sockaddr_in6);
1929         rip6src.sin6_family = AF_INET6;
1930         /* KAME hack: recover scopeid */
1931         (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1932
1933         LIST_FOREACH(in6p, &ripcbinfo.listhead, inp_list)
1934         {
1935                 if ((in6p->inp_vflag & INP_IPV6) == 0)
1936                         continue;
1937 #ifdef HAVE_NRL_INPCB
1938                 if (!(in6p->in6p_flags & INP_IPV6))
1939                         continue;
1940 #endif
1941                 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1942                         continue;
1943                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1944                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1945                         continue;
1946                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1947                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1948                         continue;
1949                 if (in6p->in6p_icmp6filt
1950                     && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1951                                  in6p->in6p_icmp6filt))
1952                         continue;
1953                 if (last) {
1954                         struct  mbuf *n;
1955                         if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1956                                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1957                                         ip6_savecontrol(last, &opts, ip6, n);
1958                                 /* strip intermediate headers */
1959                                 m_adj(n, off);
1960                                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1961                                                  (struct sockaddr *)&rip6src,
1962                                                  n, opts) == 0) {
1963                                         /* should notify about lost packet */
1964                                         m_freem(n);
1965                                         if (opts) {
1966                                                 m_freem(opts);
1967                                         }
1968                                 } else
1969                                         sorwakeup(last->in6p_socket);
1970                                 opts = NULL;
1971                         }
1972                 }
1973                 last = in6p;
1974         }
1975         if (last) {
1976                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1977                         ip6_savecontrol(last, &opts, ip6, m);
1978                 /* strip intermediate headers */
1979                 m_adj(m, off);
1980                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1981                                  (struct sockaddr *)&rip6src, m, opts) == 0) {
1982                         m_freem(m);
1983                         if (opts)
1984                                 m_freem(opts);
1985                 } else
1986                         sorwakeup(last->in6p_socket);
1987         } else {
1988                 m_freem(m);
1989                 ip6stat.ip6s_delivered--;
1990         }
1991         return IPPROTO_DONE;
1992 }
1993
1994 /*
1995  * Reflect the ip6 packet back to the source.
1996  * OFF points to the icmp6 header, counted from the top of the mbuf.
1997  */
1998 void
1999 icmp6_reflect(m, off)
2000         struct  mbuf *m;
2001         size_t off;
2002 {
2003         struct ip6_hdr *ip6;
2004         struct icmp6_hdr *icmp6;
2005         struct in6_ifaddr *ia;
2006         struct in6_addr t, *src = 0;
2007         int plen;
2008         int type, code;
2009         struct ifnet *outif = NULL;
2010         struct sockaddr_in6 sa6_src, sa6_dst;
2011 #ifdef COMPAT_RFC1885
2012         int mtu = IPV6_MMTU;
2013         struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
2014 #endif
2015
2016         /* too short to reflect */
2017         if (off < sizeof(struct ip6_hdr)) {
2018                 nd6log((LOG_DEBUG,
2019                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2020                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2021                     __FILE__, __LINE__));
2022                 goto bad;
2023         }
2024
2025         /*
2026          * If there are extra headers between IPv6 and ICMPv6, strip
2027          * off that header first.
2028          */
2029 #ifdef DIAGNOSTIC
2030         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2031                 panic("assumption failed in icmp6_reflect");
2032 #endif
2033         if (off > sizeof(struct ip6_hdr)) {
2034                 size_t l;
2035                 struct ip6_hdr nip6;
2036
2037                 l = off - sizeof(struct ip6_hdr);
2038                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2039                 m_adj(m, l);
2040                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2041                 if (m->m_len < l) {
2042                         if ((m = m_pullup(m, l)) == NULL)
2043                                 return;
2044                 }
2045                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2046         } else /* off == sizeof(struct ip6_hdr) */ {
2047                 size_t l;
2048                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2049                 if (m->m_len < l) {
2050                         if ((m = m_pullup(m, l)) == NULL)
2051                                 return;
2052                 }
2053         }
2054         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2055         ip6 = mtod(m, struct ip6_hdr *);
2056         ip6->ip6_nxt = IPPROTO_ICMPV6;
2057         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2058         type = icmp6->icmp6_type; /* keep type for statistics */
2059         code = icmp6->icmp6_code; /* ditto. */
2060
2061         t = ip6->ip6_dst;
2062         /*
2063          * ip6_input() drops a packet if its src is multicast.
2064          * So, the src is never multicast.
2065          */
2066         ip6->ip6_dst = ip6->ip6_src;
2067
2068         /*
2069          * XXX: make sure to embed scope zone information, using
2070          * already embedded IDs or the received interface (if any).
2071          * Note that rcvif may be NULL.
2072          * TODO: scoped routing case (XXX).
2073          */
2074         bzero(&sa6_src, sizeof(sa6_src));
2075         sa6_src.sin6_family = AF_INET6;
2076         sa6_src.sin6_len = sizeof(sa6_src);
2077         sa6_src.sin6_addr = ip6->ip6_dst;
2078         in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2079         in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2080         bzero(&sa6_dst, sizeof(sa6_dst));
2081         sa6_dst.sin6_family = AF_INET6;
2082         sa6_dst.sin6_len = sizeof(sa6_dst);
2083         sa6_dst.sin6_addr = t;
2084         in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2085         in6_embedscope(&t, &sa6_dst, NULL, NULL);
2086
2087 #ifdef COMPAT_RFC1885
2088         /*
2089          * xxx guess MTU
2090          * RFC 1885 requires that echo reply should be truncated if it
2091          * does not fit in with (return) path MTU, but the description was
2092          * removed in the new spec.
2093          */
2094         if (icmp6_reflect_rt.ro_rt == 0 ||
2095             ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2096                 if (icmp6_reflect_rt.ro_rt) {
2097                         RTFREE(icmp6_reflect_rt.ro_rt);
2098                         icmp6_reflect_rt.ro_rt = 0;
2099                 }
2100                 bzero(sin6, sizeof(*sin6));
2101                 sin6->sin6_family = PF_INET6;
2102                 sin6->sin6_len = sizeof(struct sockaddr_in6);
2103                 sin6->sin6_addr = ip6->ip6_dst;
2104
2105                 rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2106                             RTF_PRCLONING);
2107         }
2108
2109         if (icmp6_reflect_rt.ro_rt == 0)
2110                 goto bad;
2111
2112         if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2113             && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2114                 mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2115
2116         if (mtu < m->m_pkthdr.len) {
2117                 plen -= (m->m_pkthdr.len - mtu);
2118                 m_adj(m, mtu - m->m_pkthdr.len);
2119         }
2120 #endif
2121         /*
2122          * If the incoming packet was addressed directly to us(i.e. unicast),
2123          * use dst as the src for the reply.
2124          * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2125          * (for example) when we encounter an error while forwarding procedure
2126          * destined to a duplicated address of ours.
2127          */
2128         for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2129                 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2130                     (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2131                         src = &t;
2132                         break;
2133                 }
2134         if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2135                 /*
2136                  * This is the case if the dst is our link-local address
2137                  * and the sender is also ourselves.
2138                  */
2139                 src = &t;
2140         }
2141
2142         if (src == 0) {
2143                 int e;
2144                 struct route_in6 ro;
2145
2146                 /*
2147                  * This case matches to multicasts, our anycast, or unicasts
2148                  * that we do not own.  Select a source address based on the
2149                  * source address of the erroneous packet.
2150                  */
2151                 bzero(&ro, sizeof(ro));
2152                 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2153                 if (ro.ro_rt)
2154                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2155                 if (src == NULL) {
2156                         nd6log((LOG_DEBUG,
2157                             "icmp6_reflect: source can't be determined: "
2158                             "dst=%s, error=%d\n",
2159                             ip6_sprintf(&sa6_src.sin6_addr), e));
2160                         goto bad;
2161                 }
2162         }
2163
2164         ip6->ip6_src = *src;
2165
2166         ip6->ip6_flow = 0;
2167         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2168         ip6->ip6_vfc |= IPV6_VERSION;
2169         ip6->ip6_nxt = IPPROTO_ICMPV6;
2170         if (m->m_pkthdr.rcvif) {
2171                 /* XXX: This may not be the outgoing interface */
2172                 ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
2173         } else
2174                 ip6->ip6_hlim = ip6_defhlim;
2175
2176         icmp6->icmp6_cksum = 0;
2177         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2178                                         sizeof(struct ip6_hdr), plen);
2179
2180         /*
2181          * XXX option handling
2182          */
2183
2184         m->m_flags &= ~(M_BCAST|M_MCAST);
2185
2186 #ifdef COMPAT_RFC1885
2187         ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2188 #else
2189         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2190 #endif
2191         if (outif)
2192                 icmp6_ifoutstat_inc(outif, type, code);
2193
2194         return;
2195
2196  bad:
2197         m_freem(m);
2198         return;
2199 }
2200
2201 void
2202 icmp6_fasttimo()
2203 {
2204
2205         mld6_fasttimeo();
2206 }
2207
2208 static const char *
2209 icmp6_redirect_diag(src6, dst6, tgt6)
2210         struct in6_addr *src6;
2211         struct in6_addr *dst6;
2212         struct in6_addr *tgt6;
2213 {
2214         static char buf[1024];
2215         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2216                 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2217         return buf;
2218 }
2219
2220 void
2221 icmp6_redirect_input(m, off)
2222         struct mbuf *m;
2223         int off;
2224 {
2225         struct ifnet *ifp = m->m_pkthdr.rcvif;
2226         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2227         struct nd_redirect *nd_rd;
2228         int icmp6len = ntohs(ip6->ip6_plen);
2229         char *lladdr = NULL;
2230         int lladdrlen = 0;
2231         u_char *redirhdr = NULL;
2232         int redirhdrlen = 0;
2233         struct rtentry *rt = NULL;
2234         int is_router;
2235         int is_onlink;
2236         struct in6_addr src6 = ip6->ip6_src;
2237         struct in6_addr redtgt6;
2238         struct in6_addr reddst6;
2239         union nd_opts ndopts;
2240
2241         if (!m || !ifp)
2242                 return;
2243
2244         /* XXX if we are router, we don't update route by icmp6 redirect */
2245         if (ip6_forwarding)
2246                 goto freeit;
2247         if (!icmp6_rediraccept)
2248                 goto freeit;
2249
2250 #ifndef PULLDOWN_TEST
2251         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2252         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2253 #else
2254         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2255         if (nd_rd == NULL) {
2256                 icmp6stat.icp6s_tooshort++;
2257                 return;
2258         }
2259 #endif
2260         redtgt6 = nd_rd->nd_rd_target;
2261         reddst6 = nd_rd->nd_rd_dst;
2262
2263         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2264                 redtgt6.s6_addr16[1] = htons(ifp->if_index);
2265         if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2266                 reddst6.s6_addr16[1] = htons(ifp->if_index);
2267
2268         /* validation */
2269         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2270                 nd6log((LOG_ERR,
2271                         "ICMP6 redirect sent from %s rejected; "
2272                         "must be from linklocal\n", ip6_sprintf(&src6)));
2273                 goto bad;
2274         }
2275         if (ip6->ip6_hlim != 255) {
2276                 nd6log((LOG_ERR,
2277                         "ICMP6 redirect sent from %s rejected; "
2278                         "hlim=%d (must be 255)\n",
2279                         ip6_sprintf(&src6), ip6->ip6_hlim));
2280                 goto bad;
2281         }
2282     {
2283         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2284         struct sockaddr_in6 sin6;
2285         struct in6_addr *gw6;
2286
2287         bzero(&sin6, sizeof(sin6));
2288         sin6.sin6_family = AF_INET6;
2289         sin6.sin6_len = sizeof(struct sockaddr_in6);
2290         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2291         rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2292         if (rt) {
2293                 if (rt->rt_gateway == NULL ||
2294                     rt->rt_gateway->sa_family != AF_INET6) {
2295                         nd6log((LOG_ERR,
2296                             "ICMP6 redirect rejected; no route "
2297                             "with inet6 gateway found for redirect dst: %s\n",
2298                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2299                         RTFREE(rt);
2300                         goto bad;
2301                 }
2302
2303                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2304                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2305                         nd6log((LOG_ERR,
2306                                 "ICMP6 redirect rejected; "
2307                                 "not equal to gw-for-src=%s (must be same): "
2308                                 "%s\n",
2309                                 ip6_sprintf(gw6),
2310                                 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2311                         RTFREE(rt);
2312                         goto bad;
2313                 }
2314         } else {
2315                 nd6log((LOG_ERR,
2316                         "ICMP6 redirect rejected; "
2317                         "no route found for redirect dst: %s\n",
2318                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2319                 goto bad;
2320         }
2321         RTFREE(rt);
2322         rt = NULL;
2323     }
2324         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2325                 nd6log((LOG_ERR,
2326                         "ICMP6 redirect rejected; "
2327                         "redirect dst must be unicast: %s\n",
2328                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2329                 goto bad;
2330         }
2331
2332         is_router = is_onlink = 0;
2333         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2334                 is_router = 1;  /* router case */
2335         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2336                 is_onlink = 1;  /* on-link destination case */
2337         if (!is_router && !is_onlink) {
2338                 nd6log((LOG_ERR,
2339                         "ICMP6 redirect rejected; "
2340                         "neither router case nor onlink case: %s\n",
2341                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2342                 goto bad;
2343         }
2344         /* validation passed */
2345
2346         icmp6len -= sizeof(*nd_rd);
2347         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2348         if (nd6_options(&ndopts) < 0) {
2349                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2350                         "invalid ND option, rejected: %s\n",
2351                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2352                 /* nd6_options have incremented stats */
2353                 goto freeit;
2354         }
2355
2356         if (ndopts.nd_opts_tgt_lladdr) {
2357                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2358                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2359         }
2360
2361         if (ndopts.nd_opts_rh) {
2362                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2363                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2364         }
2365
2366         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2367                 nd6log((LOG_INFO,
2368                         "icmp6_redirect_input: lladdrlen mismatch for %s "
2369                         "(if %d, icmp6 packet %d): %s\n",
2370                         ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2371                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2372                 goto bad;
2373         }
2374
2375         /* RFC 2461 8.3 */
2376         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2377                          is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2378
2379         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2380                 /* perform rtredirect */
2381                 struct sockaddr_in6 sdst;
2382                 struct sockaddr_in6 sgw;
2383                 struct sockaddr_in6 ssrc;
2384
2385                 bzero(&sdst, sizeof(sdst));
2386                 bzero(&sgw, sizeof(sgw));
2387                 bzero(&ssrc, sizeof(ssrc));
2388                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2389                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2390                         sizeof(struct sockaddr_in6);
2391                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2392                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2393                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2394                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2395                            (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2396                            (struct sockaddr *)&ssrc,
2397                            (struct rtentry **)NULL);
2398         }
2399         /* finally update cached route in each socket via pfctlinput */
2400     {
2401         struct sockaddr_in6 sdst;
2402
2403         bzero(&sdst, sizeof(sdst));
2404         sdst.sin6_family = AF_INET6;
2405         sdst.sin6_len = sizeof(struct sockaddr_in6);
2406         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2407         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2408 #ifdef IPSEC
2409         key_sa_routechange((struct sockaddr *)&sdst);
2410 #endif
2411     }
2412
2413  freeit:
2414         m_freem(m);
2415         return;
2416
2417  bad:
2418         icmp6stat.icp6s_badredirect++;
2419         m_freem(m);
2420 }
2421
2422 void
2423 icmp6_redirect_output(m0, rt)
2424         struct mbuf *m0;
2425         struct rtentry *rt;
2426 {
2427         struct ifnet *ifp;      /* my outgoing interface */
2428         struct in6_addr *ifp_ll6;
2429         struct in6_addr *router_ll6;
2430         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2431         struct mbuf *m = NULL;  /* newly allocated one */
2432         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2433         struct nd_redirect *nd_rd;
2434         size_t maxlen;
2435         u_char *p;
2436         struct ifnet *outif = NULL;
2437         struct sockaddr_in6 src_sa;
2438
2439         icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2440
2441         /* if we are not router, we don't send icmp6 redirect */
2442         if (!ip6_forwarding || ip6_accept_rtadv)
2443                 goto fail;
2444
2445         /* sanity check */
2446         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2447                 goto fail;
2448
2449         /*
2450          * Address check:
2451          *  the source address must identify a neighbor, and
2452          *  the destination address must not be a multicast address
2453          *  [RFC 2461, sec 8.2]
2454          */
2455         sip6 = mtod(m0, struct ip6_hdr *);
2456         bzero(&src_sa, sizeof(src_sa));
2457         src_sa.sin6_family = AF_INET6;
2458         src_sa.sin6_len = sizeof(src_sa);
2459         src_sa.sin6_addr = sip6->ip6_src;
2460         /* we don't currently use sin6_scope_id, but eventually use it */
2461         src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2462         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2463                 goto fail;
2464         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2465                 goto fail;      /* what should we do here? */
2466
2467         /* rate limit */
2468         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2469                 goto fail;
2470
2471         /*
2472          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2473          * we almost always ask for an mbuf cluster for simplicity.
2474          * (MHLEN < IPV6_MMTU is almost always true)
2475          */
2476 #if IPV6_MMTU >= MCLBYTES
2477 # error assumption failed about IPV6_MMTU and MCLBYTES
2478 #endif
2479         MGETHDR(m, M_DONTWAIT, MT_HEADER);
2480         if (m && IPV6_MMTU >= MHLEN)
2481                 MCLGET(m, M_DONTWAIT);
2482         if (!m)
2483                 goto fail;
2484         m->m_pkthdr.rcvif = NULL;
2485         m->m_len = 0;
2486         maxlen = M_TRAILINGSPACE(m);
2487         maxlen = min(IPV6_MMTU, maxlen);
2488         /* just for safety */
2489         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2490             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2491                 goto fail;
2492         }
2493
2494         {
2495                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2496                 struct in6_ifaddr *ia;
2497                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2498                                                  IN6_IFF_NOTREADY|
2499                                                  IN6_IFF_ANYCAST)) == NULL)
2500                         goto fail;
2501                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2502         }
2503
2504         /* get ip6 linklocal address for the router. */
2505         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2506                 struct sockaddr_in6 *sin6;
2507                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2508                 router_ll6 = &sin6->sin6_addr;
2509                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2510                         router_ll6 = (struct in6_addr *)NULL;
2511         } else
2512                 router_ll6 = (struct in6_addr *)NULL;
2513
2514         /* ip6 */
2515         ip6 = mtod(m, struct ip6_hdr *);
2516         ip6->ip6_flow = 0;
2517         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2518         ip6->ip6_vfc |= IPV6_VERSION;
2519         /* ip6->ip6_plen will be set later */
2520         ip6->ip6_nxt = IPPROTO_ICMPV6;
2521         ip6->ip6_hlim = 255;
2522         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2523         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2524         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2525
2526         /* ND Redirect */
2527         nd_rd = (struct nd_redirect *)(ip6 + 1);
2528         nd_rd->nd_rd_type = ND_REDIRECT;
2529         nd_rd->nd_rd_code = 0;
2530         nd_rd->nd_rd_reserved = 0;
2531         if (rt->rt_flags & RTF_GATEWAY) {
2532                 /*
2533                  * nd_rd->nd_rd_target must be a link-local address in
2534                  * better router cases.
2535                  */
2536                 if (!router_ll6)
2537                         goto fail;
2538                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2539                       sizeof(nd_rd->nd_rd_target));
2540                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2541                       sizeof(nd_rd->nd_rd_dst));
2542         } else {
2543                 /* make sure redtgt == reddst */
2544                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2545                       sizeof(nd_rd->nd_rd_target));
2546                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2547                       sizeof(nd_rd->nd_rd_dst));
2548         }
2549
2550         p = (u_char *)(nd_rd + 1);
2551
2552         if (!router_ll6)
2553                 goto nolladdropt;
2554
2555     {
2556         /* target lladdr option */
2557         struct rtentry *rt_router = NULL;
2558         int len;
2559         struct sockaddr_dl *sdl;
2560         struct nd_opt_hdr *nd_opt;
2561         char *lladdr;
2562
2563         rt_router = nd6_lookup(router_ll6, 0, ifp);
2564         if (!rt_router)
2565                 goto nolladdropt;
2566         len = sizeof(*nd_opt) + ifp->if_addrlen;
2567         len = (len + 7) & ~7;   /* round by 8 */
2568         /* safety check */
2569         if (len + (p - (u_char *)ip6) > maxlen)
2570                 goto nolladdropt;
2571         if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2572             (rt_router->rt_flags & RTF_LLINFO) &&
2573             (rt_router->rt_gateway->sa_family == AF_LINK) &&
2574             (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2575             sdl->sdl_alen) {
2576                 nd_opt = (struct nd_opt_hdr *)p;
2577                 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2578                 nd_opt->nd_opt_len = len >> 3;
2579                 lladdr = (char *)(nd_opt + 1);
2580                 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2581                 p += len;
2582         }
2583     }
2584 nolladdropt:;
2585
2586         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2587
2588         /* just to be safe */
2589 #ifdef M_DECRYPTED      /*not openbsd*/
2590         if (m0->m_flags & M_DECRYPTED)
2591                 goto noredhdropt;
2592 #endif
2593         if (p - (u_char *)ip6 > maxlen)
2594                 goto noredhdropt;
2595
2596     {
2597         /* redirected header option */
2598         int len;
2599         struct nd_opt_rd_hdr *nd_opt_rh;
2600
2601         /*
2602          * compute the maximum size for icmp6 redirect header option.
2603          * XXX room for auth header?
2604          */
2605         len = maxlen - (p - (u_char *)ip6);
2606         len &= ~7;
2607
2608         /* This is just for simplicity. */
2609         if (m0->m_pkthdr.len != m0->m_len) {
2610                 if (m0->m_next) {
2611                         m_freem(m0->m_next);
2612                         m0->m_next = NULL;
2613                 }
2614                 m0->m_pkthdr.len = m0->m_len;
2615         }
2616
2617         /*
2618          * Redirected header option spec (RFC2461 4.6.3) talks nothing
2619          * about padding/truncate rule for the original IP packet.
2620          * From the discussion on IPv6imp in Feb 1999, the consensus was:
2621          * - "attach as much as possible" is the goal
2622          * - pad if not aligned (original size can be guessed by original
2623          *   ip6 header)
2624          * Following code adds the padding if it is simple enough,
2625          * and truncates if not.
2626          */
2627         if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2628                 panic("assumption failed in %s:%d", __FILE__, __LINE__);
2629
2630         if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2631                 /* not enough room, truncate */
2632                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2633         } else {
2634                 /* enough room, pad or truncate */
2635                 size_t extra;
2636
2637                 extra = m0->m_pkthdr.len % 8;
2638                 if (extra) {
2639                         /* pad if easy enough, truncate if not */
2640                         if (8 - extra <= M_TRAILINGSPACE(m0)) {
2641                                 /* pad */
2642                                 m0->m_len += (8 - extra);
2643                                 m0->m_pkthdr.len += (8 - extra);
2644                         } else {
2645                                 /* truncate */
2646                                 m0->m_pkthdr.len -= extra;
2647                                 m0->m_len -= extra;
2648                         }
2649                 }
2650                 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2651                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2652         }
2653
2654         nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2655         bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2656         nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2657         nd_opt_rh->nd_opt_rh_len = len >> 3;
2658         p += sizeof(*nd_opt_rh);
2659         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2660
2661         /* connect m0 to m */
2662         m->m_next = m0;
2663         m->m_pkthdr.len = m->m_len + m0->m_len;
2664     }
2665 noredhdropt:;
2666
2667         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2668                 sip6->ip6_src.s6_addr16[1] = 0;
2669         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2670                 sip6->ip6_dst.s6_addr16[1] = 0;
2671 #if 0
2672         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2673                 ip6->ip6_src.s6_addr16[1] = 0;
2674         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2675                 ip6->ip6_dst.s6_addr16[1] = 0;
2676 #endif
2677         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2678                 nd_rd->nd_rd_target.s6_addr16[1] = 0;
2679         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2680                 nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2681
2682         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2683
2684         nd_rd->nd_rd_cksum = 0;
2685         nd_rd->nd_rd_cksum
2686                 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2687
2688         /* send the packet to outside... */
2689         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2690         if (outif) {
2691                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2692                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2693         }
2694         icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2695
2696         return;
2697
2698 fail:
2699         if (m)
2700                 m_freem(m);
2701         if (m0)
2702                 m_freem(m0);
2703 }
2704
2705 #ifdef HAVE_NRL_INPCB
2706 #define sotoin6pcb      sotoinpcb
2707 #define in6pcb          inpcb
2708 #define in6p_icmp6filt  inp_icmp6filt
2709 #endif
2710 /*
2711  * ICMPv6 socket option processing.
2712  */
2713 int
2714 icmp6_ctloutput(so, sopt)
2715         struct socket *so;
2716         struct sockopt *sopt;
2717 {
2718         int error = 0;
2719         int optlen;
2720         struct inpcb *inp = sotoinpcb(so);
2721         int level, op, optname;
2722
2723         if (sopt) {
2724                 level = sopt->sopt_level;
2725                 op = sopt->sopt_dir;
2726                 optname = sopt->sopt_name;
2727                 optlen = sopt->sopt_valsize;
2728         } else
2729                 level = op = optname = optlen = 0;
2730
2731         if (level != IPPROTO_ICMPV6) {
2732                 return EINVAL;
2733         }
2734
2735         switch (op) {
2736         case PRCO_SETOPT:
2737                 switch (optname) {
2738                 case ICMP6_FILTER:
2739                     {
2740                         struct icmp6_filter *p;
2741
2742                         if (optlen != sizeof(*p)) {
2743                                 error = EMSGSIZE;
2744                                 break;
2745                         }
2746                         if (inp->in6p_icmp6filt == NULL) {
2747                                 error = EINVAL;
2748                                 break;
2749                         }
2750                         error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2751                                 optlen);
2752                         break;
2753                     }
2754
2755                 default:
2756                         error = ENOPROTOOPT;
2757                         break;
2758                 }
2759                 break;
2760
2761         case PRCO_GETOPT:
2762                 switch (optname) {
2763                 case ICMP6_FILTER:
2764                     {
2765                         if (inp->in6p_icmp6filt == NULL) {
2766                                 error = EINVAL;
2767                                 break;
2768                         }
2769                         error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2770                                 sizeof(struct icmp6_filter));
2771                         break;
2772                     }
2773
2774                 default:
2775                         error = ENOPROTOOPT;
2776                         break;
2777                 }
2778                 break;
2779         }
2780
2781         return(error);
2782 }
2783 #ifdef HAVE_NRL_INPCB
2784 #undef sotoin6pcb
2785 #undef in6pcb
2786 #undef in6p_icmp6filt
2787 #endif
2788
2789 #ifndef HAVE_PPSRATECHECK
2790 #ifndef timersub
2791 #define timersub(tvp, uvp, vvp)                                         \
2792         do {                                                            \
2793                 (vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;          \
2794                 (vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;       \
2795                 if ((vvp)->tv_usec < 0) {                               \
2796                         (vvp)->tv_sec--;                                \
2797                         (vvp)->tv_usec += 1000000;                      \
2798                 }                                                       \
2799         } while (0)
2800 #endif
2801
2802 /*
2803  * ppsratecheck(): packets (or events) per second limitation.
2804  */
2805 static int
2806 ppsratecheck(lasttime, curpps, maxpps)
2807         struct timeval *lasttime;
2808         int *curpps;
2809         int maxpps;     /* maximum pps allowed */
2810 {
2811         struct timeval tv, delta;
2812         int rv;
2813
2814         microtime(&tv);
2815
2816         timersub(&tv, lasttime, &delta);
2817
2818         /*
2819          * Check for 0,0 so that the message will be seen at least once.
2820          * If more than one second has passed since the last update of
2821          * lasttime, reset the counter.
2822          *
2823          * We do increment *curpps even in *curpps < maxpps case, as some may
2824          * try to use *curpps for stat purposes as well.
2825          */
2826         if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2827             delta.tv_sec >= 1) {
2828                 *lasttime = tv;
2829                 *curpps = 0;
2830                 rv = 1;
2831         } else if (maxpps < 0)
2832                 rv = 1;
2833         else if (*curpps < maxpps)
2834                 rv = 1;
2835         else
2836                 rv = 0;
2837
2838 #if 1 /* DIAGNOSTIC? */
2839         /* be careful about wrap-around */
2840         if (*curpps + 1 > *curpps)
2841                 *curpps = *curpps + 1;
2842 #else
2843         /*
2844          * assume that there's not too many calls to this function.
2845          * not sure if the assumption holds, as it depends on *caller's*
2846          * behavior, not the behavior of this function.
2847          * IMHO it is wrong to make assumption on the caller's behavior,
2848          * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2849          */
2850         *curpps = *curpps + 1;
2851 #endif
2852
2853         return (rv);
2854 }
2855 #endif
2856
2857 /*
2858  * Perform rate limit check.
2859  * Returns 0 if it is okay to send the icmp6 packet.
2860  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2861  * limitation.
2862  *
2863  * XXX per-destination/type check necessary?
2864  */
2865 static int
2866 icmp6_ratelimit(dst, type, code)
2867         const struct in6_addr *dst;     /* not used at this moment */
2868         const int type;                 /* not used at this moment */
2869         const int code;                 /* not used at this moment */
2870 {
2871         int ret;
2872
2873         ret = 0;        /* okay to send */
2874
2875         /* PPS limit */
2876         if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2877             icmp6errppslim)) {
2878                 /* The packet is subject to rate limit */
2879                 ret++;
2880         }
2881
2882         return ret;
2883 }